現今行車輔助系統已廣泛的被重視，搭載行車輔助系統的車輛往往能有效的減少車禍發生的機會，在此類系統中行車偵測扮演一個相當重要的腳色，如何達到高準確以及高效率的偵測系統會主載整個行車輔助系統的功效。目前最為廣泛被使用的行車偵測系統是透過逐步搜尋(sliding window search)的動作從截取到的影像中尋找目標車輛，逐步搜尋的方法可以提供高速的偵測以符合即時的需求，但由於硬體的發展，許多行車記錄器或鏡頭可以截取到更高畫質的影像，此也代表著逐步搜尋的範圍也會逐步的提升，而造成辨識速率的低落，欲解決此樣的困境，使用異質多核心系統來加速行車偵測系統會是一個相當不錯的方案。在此論文中，我們將一個逐步搜尋的車輛辨識系統作為研究對象，並使用異質多核心系統以及OpenCL平行語言來加速車輛辨識演算法，此外我們也整合了一個線性模組來減少逐步搜尋的尋找範圍。透過線性模組以及平行化偵測演算法，在Intel I5-2400以及AMD HD6670的平台上，我們可以對整個程式達到16.7倍的加速，對偵測核心更可達17.1倍。

Vehicle detection methods are playing an important role for driver assistance systems. Developing a high accuracy and efficiency vehicle detection system thus becomes crucial. One of the popular approaches is the scanning method which is based on the sliding window search for locating the vehicles from the input images. Such method provides a high detection rate with a time consuming process that identifies the vehicle from each sliding window. The searching time can be unacceptable as the searching space grows. This raises an optimization opportunity to exploit modern
heterogeneous multicore system to accelerate the vehicle detection process. In this paper, we present a case study to accelerate a sliding-window based vehicle detection algorithm on a heterogeneous multicore systems using OpenCL designs. Unlike transitional detection algorithm, we integrate linear model into our vehicle detection method to reduce search space. We give a detail execution profiling on each component of original vehicle detection algorithm and explore the potential parallelism. The experiment is based on a heterogeneous multicore platform that includes an Intel i5-2400 processor and a AMD HD6670 GPU. Also an Open64-based OpenCL compiler is employed to compile the cl code for the GPU. Significant performance speed-up is achieved with our parallelization and optimization, the maximum speed-up for the vehicle detection kernel and whole application is 17.1 and 16.7 respectively.

Abstract i
Contents ii
List of Figures iv
1 Introduction 1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
2 Background 4
2.1 Vehicle Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 OpenCL overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3 Algorithm Overview 8
3.1 Linear width model for reducing search space . . . . . . . . . . . . . 8
3.2 Linear Model for Vehicle Detection Algorithm . . . . . . . . . . . . . 10
4 Parallelization strategy 14
4.1 Profiling and Partitioning . . . . . . . . . . . . . . . . . . . . . . . . 14
4.2 OpenCL Vector Computations . . . . . . . . . . . . . . . . . . . . . . 17
4.3 Overlapping Execution with Double Buffering . . . . . . . . . . . . . 18
5 Experimental Results 20
5.1 Experimental environment . . . . . . . . . . . . . . . . . . . . . . . . 20
5.2 OpenCC-Open64 based OpenCL compiler . . . . . . . . . . . . . . . 21
5.3 Performance Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . 22
5.4 Accuracy analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
6 conclusion 26
References 26

[1] W. Jones, “Keeping cars from crashing,” Spectrum, IEEE, vol. 38, no. 9, pp.40–45, Sep. 2001. [Online]. Available: http://ieeexplore.ieee.org/xpls/abs all. jsp?arnumber=946636&tag=1
[2] B. E. Digregorio, “Safer driving in the dead of night [infrared vision systems],” IEEE Spectr., vol. 43, no. 3, pp. 20–21, Mar. 2006. [Online]. Available: http://dx.doi.org/10.1109/MSPEC.2006.1604835
[3] Z. Sun, G. Bebis, and R. Miller, “On-road vehicle detection: A review,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 28, pp. 694–711,2006.
[4] M. D, G. M, and G. A., “High performance and scalable gpu graph traversal.”Technical Report, No.CS-2011-05, 2011.
[5] W. Wang, T. Zhang, Y. Zhang, and H. Jia, “Parallelization and performance op- timization on face detection algorithm with opencl: A case study,” TSINGHUA SCIENCE AND TECHNOLOGY, vol. 17, no. 3, pp. 287–295, June. 2013.
[6] G. Wang, Y. Xiong, J. Yun, and J. R. Cavallaro, “Accelerating computer vision algorithms using OpenCL framework on the mobile GPU - a case study,” in IEEE 27 International Conference on Acoustics, Speech, and Signal Processing (ICASSP), May 2013.
[7] K. Komatsu, K. Sato, Y. Arai, K. Koyama, H. Takizawa, and H. Kobayashi, “Evaluating performance and portability of opencl programs,” Fifth Interna- tional Workshop on Automatic Performance Tuning, June. 2010.
[8] X. Li, Y. Gao, and Y. Liu, “Performance evaluation of fast fourier transform application on heterogeneous platforms,” International Conference on Cyber- Enabled Distributed Computing and Knowledge Discovery, 2011.
[9] F. Rattei, P. Kindt, A. Probstl, and S. Chakraborty, “Shadow-based vehicle model refinement and tracking in advanced automotive driver assistance sys- tems,” in Proceedings of the 9th IEEE Workshop on Embedded Systems for Real- time Multimedia (ESTIMedia), Taipeh, Taiwan, 2011.
[10] M. Cheon, W. Lee, C. Yoon, and M. Park, “Vision-based vehicle detection system with consideration of the detecting location.” IEEE Transactions on Intelligent Transportation Systems, vol. 13, no. 3, pp. 1243–1252, 2012. [Online]. Available: http://dblp.uni-trier.de/db/journals/tits/tits13.html#CheonLYP12
[11] J. M. Ferryman, S. J. Maybank, and A. D. Worrall, “Visual surveillance for moving vehicles,” International Journal of Computer Vision, vol. 37, no. 2, pp.187–197, 2000.
[12] N. Srinivasa, “Vision-based vehicle detection and tracking method for forward collision warning in automobiles,” IEEE In Intelligent Vehicle Symposium, vol.2002, no. 2, pp. 626–631, 2002.
[13] K. O. W. Group., “The opencl 1.2 specification,” November 2012.
[14] “Intel opencl sdk,” http://software.intel.com/en-us/vcsource/tools/opencl-sdk.
[15] “Amd opencl sdk,” http://developer.amd.com/resources/heterogeneous- computing/opencl-zone/.
[16] “Nvidia opencl sdk,” https://developer.nvidia.com/opencl.
[17] “Arm-mali opencl sdk,” http://malideveloper.arm.com/develop-for- mali/sdks/mali-opencl-sdk/.
[18] J. Lee, J. Kim, S. Seo, S. Kim, J. Park, H. Kim, T. T. Dao, Y. Cho, S. J. Seo, S. H. Lee, S. M. Cho, H. J. Song, S.-B. Suh, and J.-D. Choi, “An opencl framework for heterogeneous multicores with local memory,” in Proceedings of the 19th international conference on Parallel architectures and compilation techniques, ser. PACT ’10. New York, NY, USA: ACM, 2010, pp. 193–204. [Online]. Available: http://doi.acm.org/10.1145/1854273.1854301
[19] J.-J. Li, C.-B. Kuan, T.-Y. Wu, and J. K. Lee, “Enabling an opencl compiler for embedded multicore dsp systems,” 2012 41st International Conference on Parallel Processing Workshops, vol. 0, pp. 545–552, 2012.
[20] Y.-C. Chen, T.-F. Su, and S.-H. Lai, “Efficient vehicle detection with adaptive scan based on perspective geometry,” IEEE International Conference on Image Processing (ICIP), Sep. 2013.
[21] C.-C. Chang and C.-J. Lin, “LIBSVM: A library for support vector machines,” ACM Transactions on Intelligent Systems and Technology, vol. 2, pp. 27:1–27:27,2011.
[22] Y.-T. Lin, S.-C. Wang, W.-L. Shih, B. K.-Y. Hsieh, and J. K. Lee, “Enable opencl compiler with open64 infrastructures.” in HPCC. IEEE, 2011, pp.863–868. [Online]. Available: http://dblp.uni-trier.de/db/conf/hpcc/hpcc2011. html#LinWSHL11
[23] M. Enzweiler and D. M. Gavrila, “Monocular pedestrian detection: Survey and experiments,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 31, no. 12, pp. 2179–2195, 2009.
[24] Open64, “Online available at http://www.open64.net/,” 2013.